Zhaohui Zheng

High AP on pretrained weight FCOS_imprv_R_50_FPN_1x.pth

3

High AP on pretrained weight ```FCOS_imprv_R_50_FPN_1x.pth``` with default test command ``` python tools/test_net.py --config-file configs/fcos/fcos_imprv_R_50_FPN_1x.yaml MODEL.WEIGHT FCOS_imprv_R_50_FPN_1x.pth TEST.IMS_PER_BATCH 4 ``` ``` 2020-03-30 01:13:16,982 fcos_core.inference INFO: Start evaluation on coco_2014_minival dataset(40504...

Hi, 我最近在研究label assign，实验发现使用默认设置下的AutoAssign在VOC上的表现会低于GFL 3个点以上的AP。请问你们有在VOC上尝试过吗？我使用的是mmdet官方repo中的AutoAssign，最大epoch为4，学习率下降在第3 epoch之后，这是mmdet VOC的标准设置。GFL可取得51.8的AP，但AutoAssign只能取得48.4。请问有什么超参会对性能影响较大？

### Search before asking - [X] I have searched the YOLOv8 [issues](https://github.com/ultralytics/ultralytics/issues) and found no similar feature requests. ### Description Is there any plan of knowledge distillation on yolov8? I...

Cannot reimplement the results of CrowdDet. Why?

4

Reimplement a model in the model zoo using the provided configs `configs/crowddet/crowddet-rcnn_r50_fpn_8xb2-30e_crowdhuman.py`, but I cannot reimplement the performance of 90.0 AP. ``` 1105 10/18 17:21:55 - mmengine - INFO -...